@glennsc

Hyper-generalist; formerly: product at @unitedtalent; CTO @ CAA.

Los Angeles, CA
Joined November 2008
Glenn Scott retweeted
it's funny how scrum masters completely disappeared and no one really cared
565
579
118
12,909
1,190,509
if you find yourself in a room with enterprise sales professionals, ask them 1. what are your tips for keeping food off your clothes when you drive and eat 2. where are the best bathrooms on the local interstate and wait for the stories / arguments pro tip: avoid anything with onions in it; those are ketchup whips
33
for pacing
SITUATION DETECTED: Anthropic told investors it plans to scale its available compute to 5 gigawatts by the end of this year, up from 1.5 gigawatts last year, and to double that again to 10 gigawatts by the end of 2027, per NYT.
47
Glenn Scott retweeted
Truth is, and as corny as this may sound, a lot of Americans were pulling for this big AI moonshot of ours, taking pride in something they didn't quite understand but still taking pride. Now the companies' weasel leaders are like, hey, might be dangerous, hang on. I'm scared. Not exactly Neal Armstrong stuff. And certainly not Odysseus. They have forfeited mythologically. I'm a storyteller and they blew it.
119
172
16
1,897
36,182
one of the biggest takeaways from the @bot team’s livestream building @thursdayarena over really two days is that we would all love a chance to work with people as pleasant, smart, fearless, funny, collaborative and respectful as @mattyp @roshan_s and @poteto
30
Glenn Scott retweeted
An internal OpenAI model added this message to its persona while undergoing reinforcement learning.
100
106
91
1,958
196,279
But he isn’t pacing
After co-inventing ChatGPT, I kept asking myself: why have superhuman chat models not led to AGI? I’ve spent the last 2 years in stealth building a new way to train models (RLCD), and a new type of frontier AI model that we are releasing today: Jev • 20-200x faster • 40-400x cheaper (w/ output tokens free) • Frontier composable intelligence optimized for decisions AFAICT the shortest path to AI-based economic revolution
1
41
But sure, pace the frontier!
Cisco president @jpatel41 says cyberattacks now move in 7–9 minutes while defenders can take 45 days to patch, which is why security has to go agentic: "From the time a vulnerability is announced to the time you have an exploit happen is 7 to 9 minutes. The adversaries are already attacking a system within 7 to 9 minutes. And it's taking you 45 days to do a patch. So the current system simply does not work." "Only 20% of vulnerabilities in the world that ever get found actually get addressed and patched. 80% go unaddressed." "That's because we don't have enough people to staff up the SOCs in different companies so they can patch the vulnerabilities." "For the first time, security operation centers are gonna become more agentic. Agents will augment human capacity so you can have defense at machine scale, because these attacks are gonna happen at machine scale." @Cisco
29
“I’m told”
I’m told WH aides are advising Trump to give a national TV address in the next 72 hours, declare certain aspects of the AI race an “imminent threat to humanity,” and announce regulations pausing the things Tech companies want paused. 1)
22
Attempting human PR review after AI
1
1
127
Usage limits are the competitor’s opportunity. Your chief revenue officer is pushing you into the enemy’s hands. @claudeai limits (resets on Wednesday!) just give me an opportunity to try Cursor/Grok (excellent) and Astra (that’s next). Seriously, the switching costs are not that high. Make it easy for me not to think about it.
1
2
78
Github gettin’ spicy
Replying to @_theopowers
Next Edit Suggestions Spec-driven development / agentic loops Natural language to app Realtime multiplayer agentic development Automation A small selection of topics we were also early/first to and built since OG copilot. You don't have to like us, but playing musical chairs with your arguments to justify your antipathy doesn't lend you credibility.
1
64
Take the hard dependency on us instead!
Your software factory should not have a hard dependency on a single harness. AI SDK HarnessAgent: One API, every harness
1
2
637
Glenn Scott retweeted
YC partners next month
42
29
7
1,170
56,569
Grok @bot will make Mac Minis obtainable again
1
3
78
Hey @trq212 we need to let Fable authorize spot “token refunds” when the model admits it made an expensive mistake 😉:
2
82
Glenn Scott retweeted
11 years today since we lost Dad. Whoever said it’ll get easier with time… no the f**k it does not. #riproddypiper
1,374
1,645
79
42,569
2,042,866